Long Short Memory Process: Modeling Growth Dynamics of Microscopic Social Connectivity

How do people make friends dynamically in social networks? What are the temporal patterns for an individual increasing its social connectivity? What are the basic mechanisms governing the formation of these temporal patterns? No matter cyber or physical social systems, their structure and dynamics are mainly driven by the connectivity dynamics of each individual. However, due to the lack of empirical data, little is known about the empirical dynamic patterns of social connectivity at microscopic level, let alone the regularities or models governing these microscopic dynamics. We examine the detailed growth process of "WeChat", the largest online social network in China, with 300 million users and 4.75 billion links spanning two years. We uncover a wide range of long-term power law growth and short-term bursty growth for the social connectivity of different users. We propose three key ingredients, namely average-effect, multiscale-effect and correlation-effect, which govern the observed growth patterns at microscopic level. As a result, we propose the long short memory process incorporating these ingredients, demonstrating that it successfully reproduces the complex growth patterns observed in the empirical data. By analyzing modeling parameters, we discover statistical regularities underlying the empirical growth dynamics. Our model and discoveries provide a foundation for the microscopic mechanisms of network growth dynamics, potentially leading to implications for prediction, clustering and outlier detection on human dynamics.